KoBold Metals was valued at approximately $2.96 billion post-money following its $537 million Series C, announced in January 2025 and co-led by T. Rowe Price and Durable Capital Partners. Participants included Andreessen Horowitz Growth, Breakthrough Energy Ventures, Equinor Ventures, Mitsubishi, and Standard Investments, among others.

In November 2025, KoBold closed a $280 million Series C-III extension with participation from BOND Capital, Lightspeed Venture Partners, StepStone Group, and WCM Investment Management, alongside returning investors. A separate $200 million equity offering filed with the SEC in late 2025 had sold approximately $163.4 million as of November 2025.

Earlier rounds included a $192.5 million Series B in February 2022. Backers across the company's fundraising history include Andreessen Horowitz, Breakthrough Energy Ventures, Sam Altman's Apollo Projects, and strategic investors BHP and Mitsubishi.

KoBold Metals has raised approximately $1.21 billion in total equity financing.

KoBold Metals is a technology-enabled mineral exploration and mine-development company. Rather than selling software to miners, it uses an internal discovery system to find, evaluate, and develop deposits of copper, lithium, nickel, cobalt, and other critical minerals.

The system ingests and standardizes heterogeneous geoscience data, including historical field notes and drill logs, government geological archives, satellite imagery, geochemical assays, geophysical surveys, and LiDAR terrain models. KoBold cleans and indexes inputs across file types, coordinate systems, languages, and quality levels, allowing geologists and data scientists to query project-wide evidence without manually assembling spreadsheets and GIS workspaces.

Probabilistic and machine-learning models then predict subsurface properties, including rock type, fault orientation, conductivity, alteration patterns, and mineralized-zone grade and thickness, with explicit uncertainty estimates. The output is a ranked list of targets with quantified confidence. Teams use these rankings to select surveys, soil samples, and drill holes based on which action is expected to reduce the most relevant uncertainty.

KoBold also develops sensing hardware for its models. Hyperpod collects RGB imagery, hyperspectral data, and LiDAR at roughly ten times the speed of conventional industry approaches. Korecam360 captures continuous 360-degree records of drill core before breakage or transport, while MetaLab crushes, dries, sieves, and chemically analyzes soil samples on-site, removing weeks of wait time associated with external laboratories. KoBold also holds patents on large rigid electromagnetic induction-loop systems and compact muon-tomography detectors.

The workflow runs as a closed loop: teams select a region, clean historical data, build competing geological hypotheses, rank targets, collect new sensor data, feed results back into the system, and advance or abandon the project. KoBold's geologists, geophysicists, data scientists, software engineers, drill engineers, and laboratory personnel are the system's primary users, rather than external SaaS customers.

KoBold captures value through direct ownership of mineral assets rather than software fees. When its technology identifies a promising target, KoBold acquires or earns into the underlying mineral rights through wholly owned claims, earn-in agreements, or joint ventures, retaining equity-like exposure to the deposit. If exploration succeeds, value accrues through asset appreciation, project sales, farm-downs, retained royalties, or eventual mine-level cash flows.

The company operates through three structures. In wholly owned exploration, KoBold stakes claims, funds all fieldwork, and retains 100% of the upside. Under earn-in and joint-venture agreements, KoBold commits defined exploration spending in exchange for a percentage interest in another party's property, sharing costs and upside with partners such as BHP or state entities such as ZCCM-IH. At Mingomba, KoBold is pursuing integrated mine development, carrying the discovery through shaft construction, processing design, and infrastructure buildout. This structure gives KoBold exposure to mine-level economics while requiring multibillion-dollar capital investment and assuming construction risk.

The cost structure combines technology development with physical exploration and mining. Central technology costs, including software engineering, cloud computing, ML research, and sensor design, can be reused across 60-plus projects on four continents. Project-level costs, including drilling contractors, airborne surveys, field camps, assays, and mineral licenses, remain physical and location-specific. Mingomba's estimated $2.3 billion-plus development cost shows how the model becomes capital-intensive when KoBold retains a discovery through construction.

Go-to-market combines B2B partnerships with principal investing. Mining companies and state-owned entities contribute property access, geological data, and local infrastructure, while KoBold contributes technology, technical teams, and capital, generally in exchange for ownership rather than fees. Government relationships, including the Zambia partnership with ZCCM-IH, the DRC geological digitization agreement, and the Burundi data-digitization deal, serve as channels for project sourcing and local operating rights. There is no conventional price list. Commercial terms are bespoke and structured around earn-in spending thresholds, staged ownership milestones, free-carried state interests, and eventual production economics.

The model compounds when better data and models improve project selection, successful targets create valuable mineral interests, discoveries attract capital and partners, and additional projects generate proprietary geological data for subsequent tools. Mingomba gives KoBold a mine-development reference project when negotiating with governments, recruiting employees, and forming mining partnerships.

KoBold Metals competes in two overlapping markets: AI-enabled mineral exploration, where predictive targeting and novel sensing differentiate companies, and critical-mineral asset ownership, where mineral rights, government relationships, infrastructure access, and development execution determine outcomes.

VerAI Discoveries is KoBold's closest competitor in the asset-generation model, creating drill-ready projects and monetizing them through joint ventures and carried interests. VerAI announced a $24M Series B first close in early 2025 and has advanced more than 60 projects focused on concealed deposits in Tier 1 jurisdictions across the Americas. Its asset-light model spreads geological and jurisdictional risk more broadly than KoBold's capital-intensive mine-development approach and may appeal to mining partners seeking drill-ready targets without funding a future competitor.

Earth AI uses a similar model, combining AI targeting with proprietary drilling systems to shorten the feedback loop between prediction and physical validation. Its $20M Series B in 2025 funds exploration across copper, lithium, rare earths, and defense-related metals. Earth AI could complete more learning cycles per dollar if its smaller, faster drilling model works, though it has yet to report an asset comparable in scale to Mingomba.

Ivanhoe Electric combines proprietary hardware, machine-learning-based inversion, and mineral-asset ownership, giving it a sensor-hardware advantage. Its Typhoon system detects sulfide mineralization at depths exceeding 1.5 kilometers, while its majority-owned Computational Geosciences business converts the data into 3D subsurface models. Ivanhoe operates alliances with BHP in the United States, Ma'aden across approximately 50,000 square kilometers in Saudi Arabia, and SQM in Chile, providing access to major-miner land packages without bearing all initial exploration costs.

Fleet Space Technologies sells advanced sensing across the mining industry through ExoSphere, which combines satellite connectivity, field-deployed seismic sensors, ambient-noise tomography, and AI-generated drilling recommendations. With more than 300 surveys and over 40 customers including Rio Tinto, Barrick, and Ma'aden, Fleet could commoditize the sensing and multimodal-modeling layer that KoBold keeps proprietary. GeologicAI, which raised a $44M Series B with participation from BHP and Rio Tinto, uses core scanning, chip analysis, and drill-hole optimization to compete for workflow adoption by owning data capture at the source.

BHP is a KoBold investor, exploration partner, and strategic competitor. It operates internal AI and analytics programs, runs BHP Xplor for early-stage explorers, invests through BHP Ventures, and maintains a technology-enabled exploration alliance with Ivanhoe Electric. This coopetition gives BHP access to multiple technology suppliers, alongside proprietary global drilling, mine, and processing data that startups cannot easily reproduce.

Rio Tinto, Barrick, Glencore, and other majors are building similar capabilities through internal AI teams, partnerships with Fleet and Ideon, venture investments, and acquisitions of discoveries after junior explorers absorb early-stage risk. S&P Global data shows that recent resource additions have come disproportionately from expansions of existing deposits rather than new greenfield discoveries, making brownfield expansion around existing infrastructure a frequent alternative to greenfield exploration technology. Terra AI, whose CEO previously led AI development at KoBold, sells subsurface decision software directly to miners through a different monetization model with shared technical lineage. VRIFY's DORA platform, with more than 185 clients and ten proprietary deep-learning models, offers AI-assisted targeting to junior explorers seeking to retain their concession economics.

KoBold Metals is expanding its addressable market along three axes: moving from exploration into mine development, broadening across commodities and geographies, and building government data partnerships that create country-scale opportunities.

Mingomba marks KoBold's move from generating exploration targets to capturing mine-level economics. With more than $2 billion of planned investment and approximately 300,000 tonnes of targeted annual copper production, the project could generate multibillion-dollar annual metal revenue at KoBold's 80% non-state ownership share. If the Mingomba model works, KoBold can apply it selectively by retaining and developing tier-one discoveries while farming out smaller projects to established miners through joint ventures, earn-ins, or asset sales.

The project also creates infrastructure and operating capabilities that could improve the economics of future regional discoveries. KoBold's MOU with Africa Finance Corporation uses Mingomba copper as an anchor for the Zambia-Lobito rail corridor, potentially extending logistics access across the district.

KoBold's data infrastructure and predictive models can be applied across deposit types, providing a path into adjacent strategic materials such as tungsten, tantalum, rare earths, and by-product metals associated with copper systems. Its largest near-term commodity expansion is lithium in the DRC, where KoBold agreed to acquire AVZ Minerals' interests in the Manono lithium deposit and launched what it describes as the world's largest lithium exploration campaign in April 2026.

Lithium gives KoBold exposure to a demand cycle distinct from copper and a second potential large-scale development platform. Public-finance commitments for critical minerals in advanced economies reached approximately $65 billion in 2025, while the IEA projects a copper supply deficit of roughly 25% through 2035, creating incentives for investment in both copper and lithium. Redwood Materials and other downstream recycling and processing companies compete indirectly for roles in critical-material supply chains, but recycling alone cannot close the projected primary-supply gap.

The Burundi geological-data digitization agreement signed in March 2026 offers a repeatable market-entry strategy: KoBold helps a country convert paper archives into standardized digital datasets, then uses the resulting information base to identify investable exploration opportunities. Its DRC partnership covers large-scale exploration and geological-data digitization across the country, while programs in Australia, Finland, and Québec diversify its exposure across geological settings and sovereign risks.

Each new jurisdiction enlarges KoBold's proprietary training dataset. Discoveries and failed hypotheses across Arctic, shield, sediment-hosted, and outback environments could improve the transferability of its models and sensors, although geological differences limit how directly a model trained in one mineral belt can be applied to another.

Development concentration: Mingomba requires more than $2 billion of construction spending, will not produce copper until the early 2030s, and is KoBold's central commercial proof point, so a cost overrun, engineering delay, or copper-price downturn would damage both the project and the credibility of KoBold's broader claim that its scientific model can accelerate the full path from exploration to production.

Sovereign and title complexity: KoBold's highest-upside assets are concentrated in Zambia and the DRC, where changes to mining codes, fiscal terms, strategic-mineral classifications, export policies, and state-participation requirements can affect project economics independently of geological quality, as seen in the DRC's reclassification of lithium as a strategic substance and the unresolved ownership disputes surrounding Manono.

AI differentiation erosion: As BHP, Rio Tinto, and other majors develop internal AI capabilities and competitors such as Ivanhoe Electric, Fleet Space, VerAI, Earth AI, GeologicAI, and VRIFY offer the broader industry overlapping combinations of predictive targeting, proprietary sensors, and field validation, KoBold's technology advantage may narrow until its returns depend primarily on the quality of its mineral rights and development execution rather than its software.